AML Monitoring AI, Sovereign Transaction Analysis.

Triages suspicious transactions, maps them to FATF typologies and 6AMLD red flags, drafts SAR-ready narratives an MLRO reviews and signs. Inference runs in Intel TDX. EU company, EU-hosted customer database.

Sits alongside your existing TMS (Actimize, NetReveal, Quantexa, Hawk AI). Does not replace deterministic rules, accelerates the human triage and narrative drafting stage that consumes most AML team time.

Typical AML workflow

Where the agent fits in the AML lifecycle.

Most AML teams already have a transaction monitoring system. The bottleneck is rarely alert generation, it is alert triage, false-positive disposition and SAR narrative drafting. The agent is built for that bottleneck.

Sample prompts

What suspicious transaction analysis looks like in practice.

The agent is an AI model following a review checklist; name the FATF typologies and 6AMLD red flags you want applied. Below are simplified example prompts. Real prompts include alert payload, KYC summary, transaction graph context and historical disposition.

Analyse this customer's last 90 days of cash deposits.
Apply FATF Recommendation 10 typologies and BSA structuring tests.
Return: typology hits, confidence score, suggested SAR narrative draft.

Review this commercial invoice batch against FATF TBML red flags.
Look for over/under-invoicing, multiple invoicing, phantom shipments,
and high-risk-jurisdiction routing. Output: red-flag list + STR draft.

Cluster these 14 alerts by counterparty graph, beneficial ownership
and timing. Identify potential mule rings under 6AMLD predicate offences.
Return: cluster ranking + recommended escalation.

Regulatory context

FATF and 6AMLD, the typologies to ask for.

Prompts can draw on the FATF Recommendations and the typology guidance published in the FATF reports library, covering structuring, trade-based money laundering, professional money laundering, virtual asset service provider exposure, mule networks, sanctions evasion, terrorism financing red flags and tax-related predicates. The model works from its training and what you provide, not from a maintained typology database.

For EU regulated entities the agent can map findings against the 22 predicate offences listed under the Sixth Anti-Money Laundering Directive, including environmental crime, cybercrime and tax-fraud thresholds. Beneficial ownership opacity, PEP layering and high-risk-third-country exposure are flagged with references to the relevant 6AMLD articles, which the MLRO should check.

The agent does not interpret the law. It surfaces typology hits and red flags with references the MLRO uses as a starting point. The legal interpretation, the final classification and the SAR / STR filing decision remain human responsibilities.

Output format

SAR-ready narratives, in your jurisdiction's format.

The agent drafts the narrative section in the standard structure your reporting authority expects: subject identification, activity description, why-suspicious analysis, supporting evidence, recommended action. The MLRO edits and signs.

• FinCEN SAR, United States
• Tracfin DS, France
• NCA SAR, United Kingdom
• FIU.ITA, Italy
• FIU Spain (Sepblac), Spain
• BaFin / FIU Germany, Germany
• FINMA SR, Switzerland
• AUSTRAC SMR, Australia

Comparison

VoltageGPU AML agent vs Quantexa, ComplyAdvantage, NICE Actimize.

Quantexa, ComplyAdvantage and NICE Actimize are full transaction monitoring and screening platforms. VoltageGPU is a confidential AI agent layer designed to sit on top of one of those, not to replace it.

Pricing

Same tiers as the financial compliance pillar.

Starter $349/mo (3 seats), Pro $1,199/mo (10 seats, DeepSeek-V3.2-TEE reasoning), Enterprise $3,499/mo (unlimited seats, SCIM 2.0, RBAC admin console, admin audit log).

FAQ

Questions AML officers ask first.

Related